Generation of Local Concentration Gradients by Gas− Liquid Contacting

Apr 2, 2008 - We present a generic concept to create local concentration gradients, based on the absorption of gases or vapors in a liquid. A multilay...
0 downloads 0 Views 934KB Size
Anal. Chem. 2008, 80, 3190-3197

Generation of Local Concentration Gradients by Gas-Liquid Contacting Jorrit de Jong, Pascal W. Verheijden, Rob G. H. Lammertink,* and Matthias Wessling

Membrane Technology Group, Faculty of Science and Technology, University of Twente, P.O. Box 217, NL-7500 AE Enschede, The Netherlands

We present a generic concept to create local concentration gradients, based on the absorption of gases or vapors in a liquid. A multilayer microfluidic device with crossing gas and liquid channels is fabricated by micromilling and used to generate multiple gas-liquid contacting regions, separated by a hydrophobic membrane. Each crossing can acts as both a microdosing and microstripping region. Furthermore, the liquid and gas flow rate can be controlled independently of each other. The focus of this conceptual article is on the generation of pH gradients, by locally supplying acidic or basic gases/vapors, such as carbon dioxide, hydrochloric acid, and ammonia, visualized by pH-sensitive dyes. Stationary and moving gradients are presented in devices with 500-µm channel width, depths of 200-400 µm, and lengths of multiple centimeters. It is shown that the method allows for multiple consecutive switching gradients in a single microchannel. Absorption measurements in a microcontactor with the model system CO2/water are presented to indicate the dependence of gas absorption rate on channel depth and residence time. Achievable concentration ranges are ultimately limited by the solubility of used components. The reported devices are easy to fabricate, and their application is not limited to pH gradients. Two proof of principles are demonstrated to indicate new opportunities: (i) local crystallization of NaCl using HCl vapor and (ii) consecutive reactions of ammonia with copper(II) ions in solution. Concentration gradients are present in daily life and can appear when two or more distinct phases meet. In reaction engineering, these gradients are generally undesired, since they affect selectivity and conversion. Therefore, emphasis in research is often on rapid mixing. On the other hand, especially in analysis, concentration gradients can be beneficial. Such gradients are currently used in separation and fractionation of samples, e.g., in isoelectric focusing or conductivity gradient focusing.1 Also in the study of cell responses and in screening of crystallization conditions, concentration gradients play an important role. Three major categories can be distinguished, depending on the intended purpose: gradient in concentration of a certain * To whom correspondence should be addressed. E-mail: r.g.h.lammertink@ tnw.utwente.nl. Phone: +31 53 489 2950. Fax: +31 53 489 4611. (1) Greenlee, R. D.; Ivory, C. F. Biotechnol. Prog. 1998, 14, 300-309.

3190 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

chemical; gradient in conductivity/ionic strength; and gradient in pH. The most direct method for creating a concentration gradient in a liquid is by diffusion between laminar liquid streams with varying concentrations. Different concentration profiles can be obtained, by using layouts that can vary from simple Y junctions to complex split-and-recombine structures.2-12 In these cases, convective mass transport plays a major role, since the streams are directly linked to each other. Ismagilov and co-workers showed that the convective mass transport could be ruled out by using polycarbonate membranes as diffusional contacts between two liquids.13 Due to the high-pressure drop over the membrane, the flow rate of the liquid streams could be varied independently. Fa et al. used a similar device to selectively introduce protons into a separation channel.14 May and Hillier applied a spatially varied electric field to induce local reduction and oxidation of water into protons (at the anode) and hydroxyl ions (at the cathode).15 Disadvantage of their system was the formation of hydrogen and oxygen gases due to electrochemical reactions, leading to undesired bubble formation in the microchannels. Mitrovski and Nuzzo avoided this problem by using the oxygen reduction reaction, where oxygen is reduced to hydroxyl ions.16 Byers and co-workers immobilized the enzyme urease on a thin membrane. The ammonia that was produced in the reaction was neutralized with protons diffusing into the membrane from one side, thereby forming a pH gradient.17 A last example was (2) Jeon, N. L.; Dertinger, S. K. W.; Chiu, D. T.; Choi, I. S.; Stroock, A. D.; Whitesides, G. M. Langmuir 2000, 16, 8311-8316. (3) Dertinger, S. K. W.; Chiu, D. T.; Jeon, N. L.; Whitesides, G. M. Anal. Chem. 2001, 73, 1240-1246. (4) Holden, M. A.; Kumar, S.; Castellana, E. T.; Beskok, A.; Cremer, P. S. Sens. Actuators, B 2003, 92, 199-207. (5) Pihl, J.; Sinclair, J.; Sahlin, E.; Karlsson, M.; Petterson, F.; Olofsson, J.; Orwar, O. Anal. Chem. 2005, 77, 3897-3903. (6) Lee, J. S. H.; Hu, Y. D.; Li, D. Q. Anal. Chim. Acta 2005, 543, 99-108. (7) Chung, B. G.; Flanagan, L. A.; Rhee, S. W.; Schwartz, P. H.; Lee, A. P.; Monuki, E. S.; Jeon, N. L. Lab Chip 2005, 5, 401-406. (8) Irimia, D.; Geba, D. A.; Toner, M. Anal. Chem. 2006, 78, 3472-3477. (9) Li, C. W.; Chen, R. S.; Yang, M. S. Lab Chip 2007, 7, 1371-1373. (10) Saadi, W.; Rhee, S.; Lin, F.; Vahidi, B.; Chung, B.; Jeon, N. Biomed. Microdevices 2007, 9, 627. (11) Amarie, D.; Glazier, J. A.; Jacobson, S. C. Anal. Chem. 2007. (12) Brennen, R. A.; Yin, H.; Killeen, K. P. Anal. Chem. 2007. (13) Ismagilov, R. F.; Ng, J. M. K.; Kenis, P. A.; Whitesides, G. M. Anal. Chem. 2001, 73, 5207-5213. (14) Fa, K.; Tulock, J. J.; Sweedler, J. V.; Bohn, P. W. J. Am. Chem. Soc. 2005, 127, 13928-13933. (15) May, E. L.; Hillier, A. C. Anal. Chem. 2005, 77, 6487-6493. (16) Mitrovski, S. M.; Nuzzo, R. G. Lab Chip 2005, 5, 634-645. 10.1021/ac7023602 CCC: $40.75

© 2008 American Chemical Society Published on Web 04/02/2008

Figure 1. Schematic of a micro gas-liquid membrane contactor with multiple contacting regions: (a) 3D view; (b) top view, and (c) cross section. Local absorption of different gases or vapors leads to consecutive concentration gradients.

demonstrated by Abhyankar and co-workers, who reported a flowless system where diffusion was used to create chemical gradients.18 The methods mentioned above are based on liquid phases only. Gases or vapors can also be used to create a concentration gradient in a liquid. Beebe’s group demonstrated a pH gradient caused by absorption of acetic acid vapor.19 In previous work, we have shown the formation of a pH gradient in a porous microfluidic device by absorption of carbon dioxide through the channel walls.20,21 This gas-liquid contacting approach offers several advantages. First, the volume increase of the liquid stream is negligible, meaning no dilution of samples. Second, since gas absorption is a reversible process, the absorbed gases can be removed afterward. And third, no back diffusion of solutes from the liquid stream can occur, an issue encountered in liquid-liquid contacting. Here, we extend the approach of using a gas phase to change the composition of a liquid. Our method is based on multiple gas-liquid contacting regions, where a liquid in a microchannel is exposed to different gases or vapors. This approach enables the formation of several consecutive gradients in a single channel. A schematic is given in Figure 1. A hydrophobic membrane is integrated to obtain a stable gas-liquid interface. Using this configuration, gas and liquid flows can be regulated independently. Our concept can therefore be considered as a gas-liquid analogy to the fluid-fluid diffusional contacts that were reported by Whitesides’ group.13 The goal of this article is (a) to demonstrate the opportunities that our approach offers and (b) to give basic design and operation considerations for future development. We will provide a short background on gas-liquid contacting and show measurements of CO2 absorption in water to give an indication of the time scales involved. Subsequently, we will focus on the creation of multiple repetitive pH gradients in different device layouts and operation modes. Since many different vapors and gases can be processed, possible application areas are numerous. Besides the traditional use of gradients for analytical purposes or cell-related studies, new opportunities can also be envisioned. We demonstrate local (17) Byers, J. P.; Shah, M. B.; Fournier, R. L.; Varanasi, S. Biotechnol. Bioeng. 1993, 42, 410-420. (18) Abhyankar, V. V.; Lokuta, M. A.; Huttenlocher, A.; Beebe, D. J. Lab Chip 2006, 6, 389-393. (19) Zhao, B.; Moore, J. S.; Beebe, D. J. Science 2001, 291, 1023-1026. (20) de Jong, J.; Ankone, B.; Lammertink, R. G. H.; Wessling, M. Lab Chip 2005, 5, 1240-1247. (21) de Jong, J.; Geerken, M. J.; Lammertink, R. G. H.; Wessling, M. Chem. Eng. Technol. 2007, 30, 309-315.

crystallization of sodium chloride and local precipitation followed by dissolution of Cu2+ ions from a CuSO4 solution, using sequential reactions with NH3. BACKGROUND Membrane-assisted gas-liquid contacting is well-known in chemical engineering. Applications can be found in removal of acidic gases, such as CO2 or H2S, from exhaust gases, carbonation of soft drinks, and in blood oxygenation (artificial lungs).22 Using selective coatings, alkanes can be separated from alkenes.23 In microfluidics, membrane-based gas-liquid contacting has been applied for sensing gases, such as oxygen,16,24,25 CO2,26 NH3,27 H2S,28,29 and SO2.30 Another well-known example is the supply of oxygen through the walls of PDMS microchips, as used in cell culturing.31 However, in those examples, the obtained gradient is more an undesired consequence then a goal; much effort is put into optimizing mass transfer by mixing. In this article, the concentration gradient is the goal itself. We will first provide some general theory to be able to understand the physics in a membrane-based gas-liquid contactor and how this can be related to concentration profiles and gradients. The flux J of a gas into an absorption liquid can be expressed by eq 1:

J ) KOV∆C

(1)

Where ∆C is a concentration difference and KOV the overall masstransfer coefficient. This coefficient can be calculated using a resistances-in-series model:32 (22) Gabelman, A.; Hwang, S. T. J. Membr. Sci. 1999, 159, 61-106. (23) Nymeijer, K.; Visser, T.; Assen, R.; Wessling, M. J. Membr. Sci. 2004, 232, 107-114. (24) Wu, C.-C.; Yasukawa, T.; Shiku, H.; Matsue, T. Sens. Actuators, B 2005, 110, 342. (25) Vollmer, A. P.; Probstein, R. F.; Gilbert, R.; Thorsen, T. Lab Chip 2005, 5, 1059-1066. (26) Herber, S.; Bomer, J.; Olthuis, W.; Bergveld, P.; Berg, A.v.d. Biomed. Microdevices 2005, 7, 197. (27) Timmer, B. H.; Olthuis, W.; Van den Berg, A. Lab Chip 2004, 4, 252-255. (28) Toda, K.; Ohira, S.-I.; Ikeda, M. Anal. Chim. Acta 2004, 511, 3. (29) Ohira, S.-I.; Toda, K. Lab Chip 2005, 5, 1374-1379. (30) Ohira, S. I.; Toda, K.; Ikebe, S. I.; Dasgupta, P. K. Anal. Chem. 2002, 74, 5890-5896. (31) Walker, G. M.; Ozers, M. S.; Beebe, D. J. Biomed. Microdevices 2002, 4, 161. (32) Kreulen, H.; Smolders, C. A.; Versteeg, G. F.; Vanswaaij, W. P. M. Chem. Eng. Sci. 1993, 48, 2093-2102.

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3191

1 1 1 1 ) + + KOV kG kM mkLE

(2)

where 1/kG, 1/kM, and 1/mkLE are the mass-transfer resistances in the gas, membrane, and liquid phases, respectively. m is a physical dimensionless solubility, defined as cA,L/cA,G, with cA,L and cA,G the equilibrium concentrations of component A in the liquid and gas phases, respectively. E is a dimensionless enhancement factor that indicates the increase in absorption rate, compared to physical absorption, in case of a subsequent chemical reaction. Depending on reaction kinetics, its value can vary from 1 (no enhancement) up to values where the resistance in the liquid phase is not longer the limiting step for mass transfer. The membrane resistance 1/kM strongly depends on membrane morphology. For dense membranes, transport is determined by solubility and diffusivity,33 which are generally lower in polymers than in liquids. The membrane can thus have considerable influence on the total mass-transfer rate or even be the ratelimiting step. Therefore, in commercial gas-liquid contacting applications, mostly porous membranes are used. Most efficient operation is obtained when the gas-liquid interface is situated at the liquid side of the membrane. Transport through the membrane is then governed by diffusion in a gas phase, which is ∼10 000 times faster than diffusion in a liquid phase. For pure gases, the gas-phase resistance can even be neglected. The G-L interface will be on the liquid side as long as the pressure difference over the membrane is below the Laplace pressure, given by the Laplace equation:

∆p ) -

4γL cosθ dmax

(3)

membrane interface and the bulk liquid. The concentration at the interface ci can be related to the partial pressure pi of a gas or vapor in the pores of the membrane by Henry’s law:

p ) kc

(4)

where k is the Henry coefficient (Pa‚m3/mol). A practical guide for the application of Henry constants has been published recently.35 It is important to remark that Henry’s law is only valid for dilute solutions and does not account for subsequent partitioning or reactions of the dissolved gas within the liquid. When using basic and acidic vapors to create pH gradients, a reaction is occurring by definition: absorption leads to the formation of either protons or hydroxyl ions. The use of Henry constants for calculations in such systems can lead to underestimation of the amount of absorbed species and compensation is required. Modeling and prediction of exact concentrations therefore rapidly grows in complexity, especially when using multicomponent mixtures. Another complicating factor is that equilibrium conditions are assumed, although these may not be reached in dynamic systems. Still, basic thermodynamic data can be used to indicate the limits of the system. The maximum achievable concentration in the liquid microchannel is limited by the solubility of the gas or vapor compound at a certain partial pressure in the used liquid. Data for a large range of compounds can be found in chemical engineering handbooks, such as Perry.36 The time needed to reach the equilibrium concentration ci,max at a certain position can be estimated using Fick’s second law:

∂2ci ∂ci ) Di 2 ∂t ∂x

(5)

where γL is the surface tension of the liquid, θ the contact angle of the liquid on the membrane material, and dmax the maximum pore diameter in the membrane. When the contact angle is below 90°, the Laplace pressure becomes negative and the membrane will be directly wetted, unless a counterbalancing gas pressure is applied. From eq 3, it is clear that the choice of the membrane material and pore size is crucial to avoid wetting.34 In general, hydrophobic materials are used, such as polypropylene and Teflon, and applied membranes have pore sizes in the submicrometer region. The Laplace pressure for water in such systems (with γwater ) 72 × 10-3 N/m, θ ) 120°) is typically in the range of a few bars. It is important to remark that for small systems, such as microfluidic devices, the pressure drop over a channel can easily exceed the Laplace pressure. In that case, the application of a thin dense polymeric coating may be considered, at the cost of increased mass-transfer resistance. The application of dense coatings on the membrane, e.g., PDMS, has the additional advantage that high gas pressures can be used without bubble formation in the liquid channel. Furthermore, it enables the use of liquids with contact angles below 90° and the exploitation of selectivity of the coating material.22 Once a gas or vapor has been absorbed in the liquid phase, it will diffuse due to a concentration difference between the liquid-

where ci,0 is the concentration at t ) 0. From the analytical solution, it is evident that it takes infinite time to reach exactly the same concentration at every position. To reduce the time to reach a certain concentration, one can (a) reduce the diffusion distance, by making shallow channels, or (b) use gas/vapor streams with partial pressures that correspond to a liquid equilibrium concentration that is much higher than the required concentration, in order to have a high driving force. Furthermore, micromixers may be used to enhance mass transfer in the liquid phase. In this article, we will focus on three compounds to generate pH gradients: carbon dioxide, hydrochloric acid, and ammonia. Carbon dioxide is directly supplied as a gas, while NH3 and HCl are obtained from the vapor of solutions of these compounds. For pure CO2, the solubility in water is 0.0336 mol/L at 1 bar and 298 K, corresponding to a minimum pH of 3.9.36 The achievable pH for HCl and NH3 depends on the composition of the solutions from which the vapor is generated. The pH of these solutions

(33) Wijmans, J. G.; Baker, R. W. J. Membr. Sci. 1995, 107, 1. (34) Mavroudi, M.; Kaldis, S. P.; Sakellaropoulos, G. P. J. Membr. Sci. 2006, 272, 103.

(35) Smith, F. L.; Harvey, A. H. Chem. Eng. Prog. 2007, 103, 33-38. (36) Perry’s Chemical Engineers’ Handbook. 6th ed,; Green, D. W., Ed.; McGrawHill: New York, 1984.

3192

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

with the analytical solution

ci,max - ci(x,t) ) erf ci,max - ci,0

(x ) x

4Dit

(6)

determines the maximum pH change that can be realized. To give a rough indication, we tested the change in pH when a piece of wetted pH paper was held above solutions of 37 wt % HCl or 25 wt % NH3. Within seconds, the pH dropped to 1 (HCl) or increased to 10 (NH3), which means that a large part of the total pH range can be covered using these vapors. EXPERIMENTAL SECTION Materials. Devices were fabricated from 1-cm-thick PMMA plates. Hydrophobic polypropylene membranes (Accurel, porosity 70%) were supplied by Membrana. Demineralized water was made using a Millipore Q-water apparatus. A universal pH indicator solution was prepared by dissolving 30.6 mg of bromothymol blue, 2.6 mg of thymol blue, and 6.2 mg of methyl red in 100 mL of Q-water. In standard procedures for pH indicator solutions, often alcohol is added to enhance the dissolution. Since this compound promotes wetting of the membrane, it was left out in our experiments. Furthermore, the used concentration was relatively high compared to standard procedures in order to obtain good contrast in the shallow channels. The pH of the indicator solutions was adjusted with 0.1 M NaOH solution or 0.1 M HCl. Solutions were sieved through 0.45-µm filters (Spartan, Whatman) before testing. Carbon dioxide (99,95% Hoek Loos) and nitrogen (99.99% Hoek Loos) were used as contacting gases. Nitrogen was bubbled through either an ammonia solution (5 wt %, prepared from 25 wt % ammonia, Merck) or HCl solution (1 M, prepared from 37 wt % fuming HCl, Merck) to obtain a stream of ammonia or HCl vapor, respectively. Chip Fabrication and Experimental Setup. Microchannels with a depth of 300 µm were milled in PMMA with a Ø 0.5-mm double cutter using a Sherline 5410 automated CNC-mill. Holes of Ø 1 mm were drilled from the channel end until half the thickness of the plate. Then 1/16-in. access holes were drilled from the side. PEEK tubing (inner diameter 0.5 mm, outer diameter 1/ in, Upchurch) was glued in using Araldite two-component glue 16 and used for connections. A piece of Accurel membrane was cut to the right dimensions and directly clamped between two structured plates, using screws. Analysis by scanning electron microscopy revealed that the morphologies of the two surfaces were different, although the cross section structure of this membrane looked symmetric with interconnected pores. In our case, the most regular side with the smallest pore size and pore size distribution was facing the liquid channel. A syringe pump (Harvard 11 Picoplus) with 1-mL glass syringes (Hamilton TLL) was used for fluid flow. Nitrogen and carbon dioxide flows were controlled using Bronkhorst EL-FLOW 0-1 mL/min mass flow controllers. Small vapor generators were prepared in-house and directly connected to the gas inlet of the microfluidic devices. Characterization. The prepared microcontactors were characterized in two flow modes, with regards to the liquid side: stopped-flow and continuous flow. In both cases, gases were supplied continuously. Both single and dual gas experiments were performed. The pH gradients created by absorption and subsequent reaction of the supplied gases were visualized by pH indicator solution. Although this method does not give very strong quantitative information, it is very powerful to demonstrate the shape of concentration profiles. The relation between color and pH was determined by titration. For pH values below 4.5, the color was yellow-orange, changing via green (pH 7) to blue (pH >8.5).

Figure 2. Influence of liquid residence time on CO2 absorption in water for different liquid channel depths. Gas channel depth was 400 µm, and gas flow rate was 200 µL/min, corresponding to a residence time of 16 s. The lines indicate the results from simulations using a 2D COMSOL model.

We designed the experimental conditions in such a way that resulting pH profiles matched with the color change trajectory of the pH indicator. A Canon IXUS 6.0 digital camera was used for normal imaging. For closer optical inspection, a Zeiss Axiovert 40 optical microscope was used with a polarizer, to avoid reflections of the PMMA. To give an indication of the time scale of absorption, absorption of CO2 in Q-water at room temperature was studied as a model system. Experiments were carried out in a 20-cm-long serpentine channel of 500-µm width, in which there was continuous contact between gas and liquid sides. The channel depth of the liquid side, which is a measure for diffusion distance, was varied from 200 to 400 µm, leading to internal volumes of 20-40 µL. The gas channel had a depth of 400 µm. Pure CO2 was supplied at 1 bar with a mass flow controller at a constant gas flow rate of 200 µL/min. The conductivity at the exit was measured using the conductivity sensor of an AKTA Prime (Amersham Biosciences) and correlated to the CO2 concentration via a calibration curve. Liquid residence times were varied by applying liquid flow rates from 10 to 200 µL/min. RESULTS AND DISCUSSION Gas Absorption Rate. An important factor in the application of gas-liquid contacting is the rate of absorption of the gaseous components; ultimately, this rate determines the concentration ranges and spatial gradients that can be realized. The absorption rate depends first on interactions between the chosen gases, vapors, and absorption liquids, expressed as diffusion coefficients and solubility. Second, process parameters play a role, such as flow rates, contactor dimensions, contacting times, and applied partial pressures. Results for different liquid residence times and channel depths are shown in Figure 2 for a CO2-water system. The presented data clearly illustrate the dependence of the obtained gas concentration in the liquid on liquid residence time and diffusion distance. For longer residence times, the concentration converges to the solubility value of 0.0336 mol/L CO2 at 1 bar in water at 298 K. Fluxes of CO2 calculated from these graphs were in the order of 10-3 mol/m2‚s for the lowest residence time and channel depth, whereas longer contacting times and higher Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3193

Figure 3. Stopped-flow pH gradient. Ammonia vapor is flowing from right to left in a back structure, indicated by the dashed rectangle, at 50 µL/min. Since ammonia is absorbed, the concentration in the vapor phase decreases from right to left, leading to a lowered driving force and hence to plugs of different length (A). When the liquid in the device is shortly pumped, the plugs can be shifted (B), and new plugs can be formed (C). Starting pH value 3.5, blue indicates pH >8.5.

Figure 4. Continuous-flow pH gradient. Ammonia vapor is flowing from right to left in a back structure, indicated by the dashed rectangle, at 50 µL/min. Fluid flows from left to right with different flow rates: (a) 20, (b) 10, and (c) 5 µL/min. Starting pH value: 3.5, blue indicates pH >8.5.

channel depths led to lower values. Since a pure gas and a porous membrane were used, the main mass-transfer resistance was located in the liquid phase.34 This phase was subsequently modeled in 2D in COMSOL Multiphysics, assuming laminar flow and no slip on the walls. The concentration at the membrane interface was set to the solubility value of CO2. Average CO2 concentrations were calculated by boundary integration over the outflow boundary. The resulting trends agree reasonably with the obtained data, showing that mass transfer in gas-liquid contacting in these microfluidic devices can be accurately described. We will now focus on specific opportunities that microfluidic formats can provide in terms of local gas-liquid contacting. Single Gas Experiments. Figure 3 shows pH gradients created in a stopped-flow microfluidic contactor with multiple contacting regions. Using this flow mode, the fluid in the liquid channel can be locally saturated with a certain gas in such a way that plugs are formed. During movement of the plugs, a parabolic flow profile can be observed. The dispersion generated by the parabolic flow profile leads to different pH profiles. When one gas channel is used for multiple contacting regions, at every position a certain amount of gas will be absorbed. If a diluted gas is used, the gas concentration will decrease in the length of the gas channel, leading to a decreased driving force for transport, and ultimately to different plug lengths in the liquid channel. Such behavior can indeed be observed in the images. When the same device is operated in continuous flow, a system is obtained that is more similar to traditional gas-liquid membrane 3194

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

contactors. The main difference is that there is no continuous contact between gas and liquid side, and therefore, the pH profile will change in steps rather than in a smooth manner. A typical example is given in Figure 4. At every passing over the gasliquid contacting area, ammonia is absorbed, leading to a stepwise increase in pH. More freedom in operation can be achieved by using multiple gases. Multiple Gas Experiments. Figure 5 shows images of a dualgas chip operated in continuous mode. The meandering liquid channel is placed on alternating channels, where either NH3 vapor or CO2 is supplied. Effectively, the liquid is contacted with basic and acidic environments, thereby creating stagnant reversing pH gradients in a continuous flow, which we call “pH switch”. The flow profile in this device is illustrated by the shape of the color change trajectory. This effect enables us to check if (a) the fluid in the channel is stagnant or flowing and (b) the outlining between gas and liquid side is accurate. Besides the ratio and flow rate of the gas streams, the flow rate of the liquid stream can also be used to manipulate the pH profile, as is illustrated in Figure 6. Here we have used HCl instead of CO2. Since HCl is a much stronger acid, lower pH values can be reached, as is demonstrated by the images. The absorption of HCl/NH3 and CO2/NH3 in water, leads to the formation of a buffered solution as more and more ions are formed. Furthermore, the longer the contact time, the more the concentration of the components approaches the equilibrium value, and the lower the driving force for transport. Therefore,

Figure 5. Continuous switching pH gradients induced by pure CO2 and NH3 vapor: (a) schematic showing the back structure with the gas channels and the flow directions; (b) pH profiles for different CO2/NH3 flow ratios. The flow rate of the pH indicator was 4 µL/min for all cases, with a starting pH value of 7. Yellow pH ∼5.5; blue pH >8.5.

Figure 6. Variations of pH profile with liquid flow rate, using HCl and NH3 vapor: (a) continuous pH switch, at increasing flow rate of the pH indicator solution (HCl vapor 500 µL/min, NH3 vapor 500 µL/min); (b) stopped-flow pH switch after different contact times (HCl vapor 1000 µL/min, NH3 vapor 100 µL/min). Yellow-orange pH 8.5.

the pH gradient in continuous mode is not switching between fixed values after each repeating structure. Depending on the rate of absorption and the created buffering strength, the solution might already be stable after a short contacting period. In the examples shown thus far, the pH effect of “gas 1” was compensated by an opposite effect of “gas 2”, which led to an increase in ionic strength. A different approach is to exploit the reversibility of absorption: gas 1 can be removed with an inert stripping gas. An example is presented in Figure 7. Absorption of carbon dioxide leads to the formation of carbonic acid, which acidifies the water. This reaction is a thermodynamic equilibrium, depending on the carbon dioxide partial pressure pCO2. Since the

pCO2 is zero in the nitrogen section, CO2 is desorbed here, leading to (partial) restoration of the pH toward the initial value. This desorption process is much slower than the absorption process, since the driving force is much lower. Furthermore, the subsequent reactions of CO2 in water need to be reversed before desorption can take place. Therefore, it is not surprising that the change in pH is limited in continuous flow mode. Optimization of the residence times in the desorption section is required to increase the effectiveness. From Figures 5-7, some more observations can be made about the position of color change. We have noticed that this position can shift along the channels, even to zones where it would Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3195

Figure 7. Creation of pH gradients using absorption and desorption. Absorption of CO2 leads to acidification of the water. Nitrogen is used to shift the equilibrium and to strip the CO2 again, (partly) restoring the pH to the initial value. (a) Continuous flow mode; (b) stopped-flow mode. The pH of the starting solution was 9. Yellow pH ∼5.5; blue pH >8.5. CO2 flow rate 100 µL/min; nitrogen 1000 µL/min.

Figure 8. Demonstration of planar diffusion through the membrane: (a) schematic cross section, showing the direction of transport through the embossed porous membrane; (b) continuous-flow gradient in a device where gas and liquid channels are present in the same plane; (c) schematic of a device where crossover of gases is avoided. The liquid channel is meandering through the plate, and two separate membranes are used for gas-liquid contacting.

not be expected. Two effects play a role. First, the observed color is largely determined by the darkest color. As the gas is supplied from the bottom, a concentration gradient will be created along the depth of the channel. Mixing by diffusion in perpendicular direction causes leveling of the concentration gradient and can lead to a distinct color change. A second factor, which is in our opinion more important, is interdiffusion of the gases between the channels, in the planar direction of the membrane. In the presented experiments, the membrane was hand-tight clamped by screws, leading to negative embossing of the structure of the channels in the membrane. Starting out with a thickness of 90 µm for the fresh membrane, an average thickness of 40 µm was measured after clamping. The thickness of the freestanding membrane part between the channels remained ∼90 µm. However, with an initial porosity of 70%, still porosity was left in the compressed parts. Since there is a clear concentration gradient between the channels for both gases in our devices, diffusion will occur. In case of different pressure drops over the gas channels, even convective flux of gases may be observed between channels. Practically this means that a certain gas concentration gradient is created in the membrane, which is then superposed onto the liquid channel. The strong crossover effect is intrinsic to the use of gases; in liquid-liquid contacting, its effect is orders of magnitude lower due to higher viscous forces and lower diffusion coefficients. Transport of gases in planar direction was checked by fabricating a single plate device containing both gas and liquid channels, as is illustrated in Figure 8. Indeed, pH gradients could be observed, confirming our hypothesis. Although the diffusion effect can simplify the fabrication process, since only a single plate 3196 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

needs structuring, it may not be desired in every application. A different device layout, as depicted in Figure 8c, can offer better control. Also the use of stamping techniques to fill the porosity of the membrane except at the G-L crossings may be useful.37 Another possibility is the integration of track-etched membranes, which are commonly used in microfluidics. The pores in these membranes are oriented perpendicular to the surface, thereby avoiding diffusion in planar direction. However, it must be remarked that most track-etched membranes consist of a hydrophilic material, such as polycarbonate and polyester, and thus hydrophobization of the surface is required to prevent wetting. Recently, a review was published on the integration of membrane features in microfluidic devices.38 We strongly believe that the discussed fabrication techniques can be used for further optimization and miniaturization of the presented concept. CONCLUSIONS AND OUTLOOK In this article, we have proposed a new generic approach to generate concentration gradients in microfluidic devices. The method is based on multiple gas-liquid contacting points where a liquid in a microchannel is exposed to different gases or vapors. The contacting regions can be used for both supply and removal of gases and volatile components. We have demonstrated proofof-principles by showing pH gradients in water in different operating modes. Furthermore, we have discussed the physical (37) Chueh, B.; Huh, D.; Kyrtsos, C. R.; Houssin, T.; Futai, N.; Takayama, S. Anal. Chem. 2007, 79, 3504-3508. (38) de Jong, J.; Lammertink, R. G. H.; Wessling, M. Lab Chip 2006, 6, 11251139.

Figure 9. Effects of gas absorption on solubility of other compounds: (a) precipitation of pH indicator in continuous flow, by absorption of HCl vapor from a 37 wt % solution; (b) local crystallization in a stagnant 2 M NaCl solution, by absorption of HCl vapor from a 37 wt % solution. The dashed lines indicate the gas channel; (c) precipitation of Cu(OH)2 from the reaction of Cu2+ ions with hydroxyl ions, generated by absorption of ammonia. In excess of ammonia, a consecutive reaction leads to dissolution and complexation of the light blue Cu(OH)2 to Cu(NH3)42+, which is a dark blue complex. The fluid is stagnant, and NH3 and HCl vapor are supplied in neighboring channels. Channel width is 500 µm for all images.

processes that take place and have given practical guidelines for future development, including design considerations. Explorative experiments with CO2 absorption indicated the dependence of local concentrations on diffusion distance and diffusion time, as was expected from general theory. The results from modeling of the liquid phase demonstrated the predictability for this model system. The reported microcontactors are very easy to fabricate, and we believe that they can be useful as study tools in different areas. As pH gradient generators, they may be directly used in analysis methods or cell studies. In biotechnology, basic or acidic gases may be used to regulate pH in fermentation experiments. Currently, ammonia is already used in such systems, but supplied as a liquid.39 In reaction systems, pH-dependent reactions may be locally promoted or quenched. However, the concept of micro gas-liquid contacting using different gases can be applied more generally. The goal does not necessarily have to be a concentration gradient itself; neither is it limited to pH effects. Gas or vapor flows can be used to locally supply reagents and remove formed products. In reactor systems with sequential reactions, such a switch may lead to improved selectivity. The absorption and reactions of compounds like NH3 and CO2 lead to an effective increase in ionic strength and conductivity that can be exploited and offer the possibility to buffer a solution without dilution. Since

these compounds behave as volatile electrolytes, they can be removed afterward.40 Furthermore, the solubility of species in the liquid stream can be changed by the absorption of gases or vapors, leading to crystallization or salting-out. In fact, this last concept was triggered by the observation of small red crystals in the pH indicator experiments when using concentrated HCl vapor. Figure 9 shows a close-up of a channel in such an experiment, together with two conceptual applications: (i) local crystallization of NaCl and (ii) precipitation followed by dissolution of Cu2+ ions from a CuSO4 solution, using sequential reactions with NH3. In this last example, solids can be locally created and redissolved in a switchlike configuration. The principle of local precipitation by absorption may be applicable for local patterning. To conclude, we believe that our method extends the toolbox of microfluidics and can lead to fascinating new applications.

(39) Lee, H. L. T.; Boccazzi, P.; Ram, R. J.; Sinskey, A. J. Lab Chip 2006, 6, 1229-1235. (40) Watanabe, E. O.; Pessoa, P. D.; Miranda, E. A.; Mohamed, R. S. Biochem. Eng. J. 2006, 30, 124-129.

Received for review November 16, 2007. Accepted January 28, 2008.

ACKNOWLEDGMENT This research was carried out within the Dutch initiative “Process on a Chip” (PoaC), part of the framework Advanced Chemical Technologies for Sustainability (ACTS). We acknowledge the Dutch organization for scientific research (NWO) for financial support.

AC7023602

Analytical Chemistry, Vol. 80, No. 9, May 1, 2008

3197